the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Aug 2023
Evaluating marine dust records as templates for optical dating of Oldest Ice
Abstract. The continuous ice core record extends 800,000 years into the past, covering the period of 100,000-year glacial cycles, but not the transition from 40,000-year glacial cycles (the Mid-Pleistocene Transition, 1.2–0.7 million years ago). A primary goal of the International Partnership in Ice Core Sciences is therefore to retrieve a 1.5-million-year-old continuous ice core, increasing our understanding of this major change in the climate system and thus of fundamental climate forcings and feedbacks. However, complex glacial processes, limited bedrock data, and surprisingly young basal ice in previous cores necessitate careful reconnaissance studies before extracting a full core.
Ice borehole optical logging reflects the ice dust content and may be used to date ice quickly and inexpensively if a reference record is known. Here we explore the relationship between ice dust records and well-dated marine dust records from sediment cores in the southern Atlantic and Pacific Oceans, which lie along paths of dust sources to Antarctica. We evaluate how representative these records are of Antarctic dust both through the existing ice core record and during the older target age range, suggesting that a newly published 1.5 million year record from site U1537 near South America is likely the most robust predictor of the Oldest Ice dust signal. We then assess procedures for rapid dating of potential Oldest Ice sites, noting that the ability to detect dating errors is an essential feature. We emphasize that ongoing efforts to identify, recover, date, and interpret an Oldest Ice core should use care to avoid unfounded assumptions about the 40 kyr world based on the 100 kyr world.
- Preprint
(2969 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2023-1342', Eric Wolff, 15 Aug 2023
-
AC1: 'Reply on RC1', Jessica Ng, 09 Nov 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1342/egusphere-2023-1342-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Jessica Ng, 09 Nov 2023
-
RC2: 'Comment on egusphere-2023-1342', Frédéric Parrenin, 23 Aug 2023
In this manuscript, Jessica Ng et al. present a new marine dust record near South America which could be used as a target to date the forthcoming OldestIce ice core dust records from Antarctica. They show that their new U1537 marine dust record differs significantly from the published ODP 1090 dust record for the MPT and pre-MPT periods, 1.5-0.8 Myr ago. They argue their new site is better since the correlation with the LR04 stack stays more or less constant in the pre-MPT period with respect to the post-MPT period, contrary to the ODP 1090 record. They argue that it is possible to measure the dust record in Antarctica by logging the ice borehole, and they present such a new borehole dust logging from EPICA Dome C. They then show simple tuning strategies to quickly come-up with a time scale once the new ice dust record is available.
The paper is generally focused and straightforward, yet important, so it was a pleasure for me to review it.
I find the new dust record from U1537 very interesting. We can discuss if it is really better than ODP1090 (personnally I am quite convinced), but at the very least it is an alternative record which shows that ODP1090 should be taken with caution.
I do agree with Eric Wolff that the new borehole dust record from EPICA Dome C could get a bit more attention if it is really the first time such record is published. A quick comparison with the Coulter and lazer records would be interesting.
Regarding the end of the manuscript with the tuning strategy, I think there should be more powerful strategies, this is only a first-step strategy (but I think the authors are honnests in presenting it this way). And I do agree with Eric Wolff that this part is not as well presented and straightforward as the other parts. For example, the authors use a Nye ice flow model, but never explicitely describe the parameters they used, while for example the melting is a primary parameter for dating the old section of an ice core.
I would suggest to use the 1D model used in, e.g., Parrenin et al. (TC, 2017), Lilien et al. (TC, 2021) and Chung et al. (TC, in press), which uses a Lliboutry velocity profile. This model has an analytical thinning function and accounts for temporal variations of accumulation through a simple change of time variable. It should give a far better accuracy, while still being very easy to implement. I do not make a strong requirement to use this model, but I think it would be an improvement and I can provide guidance on request.
Moreover, the strategy is presented as decoupled between the modelling and the tuning, while I think both should be coupled: one first tune the top part, then apply the model with these dating constrains to extrapolate, then one tune the following part, etc. I personnally think the best approach would be to adjust the glaciological parameters of an ice core dating model in a Bayesian code like IceChrono/Paleochrono so as to optimize the fit with several targets, using a powerful MonteCarlo approach. (Such a method has been presented in the PhD of Jai Beeman in 2019, but unfortunately it has never been published elsewhere).
Minor comments:
- L. 40-45: The discussion of gradual vs abrupt MPT, as presented in Legrain et al. (2023) would fit nicely in this introduction, but I let you decide.
- L. ~50: In my opinion, the best evaluation of the age and state of basal ice in the Dome C area is from Chung et al. (TC, in press), but I let you decide if you want to cite it.
- L. 96: If I am correct, DFO2006 is the O2/N2 age scale of the first Dome Fuji ice core, DF1. The age scale from Dome Fuji members (2017) is DFO2006, then extended using AICC2012 for DF2. Please check, I don't think this age scale has a proper name, but I would call it DFO2006+AICC2012.
- L. ~120: In Table 1, the correlation coefficients are given for the ice core dust records, not for their log. The coeffs for the log-records are given a bit later in the manuscript, but not in Table 1. I personnally think the coeffs for the log-records are more relevant than for the raw records and I would put them in Table 1.
Citation: https://doi.org/10.5194/egusphere-2023-1342-RC2 -
AC3: 'Reply on RC2', Jessica Ng, 09 Nov 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1342/egusphere-2023-1342-AC3-supplement.pdf
-
RC3: 'Comment on egusphere-2023-1342', Lorraine Lisiecki, 02 Oct 2023
This manuscript is quite practically focused on initial assessment of drilling locations for oldest ice, proposing and demonstrating that borehole optical logs of dust would be an effective method to estimate the age of ice back to 1.5 Ma. Overall, I find this to be a valuable scientific contribution worthy of publication. However, a few portions of the manuscript need to be clarified and a few more calculations/experiments would also be very useful.
Needed clarifications:
- Line 117-118: The meaning here is ambiguous about what the authors did to “correlate” the records and put them “on a common timescale.” I think the intended meaning is that the records were sampled at common time steps in order to calculate the correlation coefficient between the signals. However, the phrasing could alternatively be interpreted as *stratigraphic* correlation that aligned the records to one common target timescale. Please clarify. If the records are all staying on their respective published timescales described in section 2, then a brief discussion of the extent to which these timescales are expected to agree or differ is warranted.
- Like Eric Wolff, I found Figure 5 to be extremely confusing. The caption, axes labels, and legend don’t seem to agree with one another. I think the blue line color was used for two different types of data. If so, the problem could be fixed by changing the line color for laser dust in panel c to another color that isn’t already used for something else.
- The description of how the artificial dust record was created is not clear, particularly “scaling the smoothed record by random factors between 0.4 - 0.7 linearly interpolated between 500 kyr intervals” (line 186). An additional concern I have about this methodology is that the laser dust record for EDC shows a much weaker dust amplitude from 500-800 ka than 0-500 ka. Are depth-dependent processes contributing to the weakening of the dust signal in the ice? If so, it would be more realistic if your artificial borehole dust record progressively decreased in amplitude farther back in time rather than scaling randomly through time.
- It is very difficult to visually discern misalignment in the DTW results, which strengthens the manuscript’s conclusion that DTW is not a suitable method for initial comparison between the borehole optical dust logs and the marine record. This point would be strengthened if the authors could find an additional way to illustrate the misalignment. For example, at the top of panel c, the artificial dust record could be plotted on its original, correct age model and arrows could be drawn to DTW results to show how dust peaks were misaligned.
Additional calculations/experiments:
- The weaker correlation for derivatives of the log of the data than the log itself is worthy of a bit more discussion/exploration. I wonder whether this might be an artifact of small differences in the age models for each record. The derivative has higher frequency variability and is therefore more likely to be affected by small age model differences/misalignments. This might actually make the derivative more useful for alignment than the log itself. The correlation of the derivative may also be advantageous if it is less sensitive to long-term trends in the mean or amplitude of the signal through time. This could potentially be evaluated by improving the alignment between the records or by analyzing synthetic records.
- Was the alignment experiment for an artificial dust signal repeated multiple times with different random noise or only once? Multiple iterations would be useful for characterizing the probability of incorrect conclusions. Additionally, what if the correlation between UF1537 and dust ice were slightly lower before 800 ka (e.g., R=0.7 instead of 0.76)?
- Eric Wolff makes a good point about the need to detect ice disturbances that could produce repeated or jumbled sections of the dust time series. I recommend creating an artificial signal that contains jumbled or duplicated sections (from the modified marine dust record) to determine whether your proposed method would detect a misfit between the simulated disturbed ice and the original marine dust record.
Citation: https://doi.org/10.5194/egusphere-2023-1342-RC3 -
AC2: 'Reply on RC3', Jessica Ng, 09 Nov 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1342/egusphere-2023-1342-AC2-supplement.pdf
Status: closed
-
RC1: 'Comment on egusphere-2023-1342', Eric Wolff, 15 Aug 2023
-
AC1: 'Reply on RC1', Jessica Ng, 09 Nov 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1342/egusphere-2023-1342-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Jessica Ng, 09 Nov 2023
-
RC2: 'Comment on egusphere-2023-1342', Frédéric Parrenin, 23 Aug 2023
In this manuscript, Jessica Ng et al. present a new marine dust record near South America which could be used as a target to date the forthcoming OldestIce ice core dust records from Antarctica. They show that their new U1537 marine dust record differs significantly from the published ODP 1090 dust record for the MPT and pre-MPT periods, 1.5-0.8 Myr ago. They argue their new site is better since the correlation with the LR04 stack stays more or less constant in the pre-MPT period with respect to the post-MPT period, contrary to the ODP 1090 record. They argue that it is possible to measure the dust record in Antarctica by logging the ice borehole, and they present such a new borehole dust logging from EPICA Dome C. They then show simple tuning strategies to quickly come-up with a time scale once the new ice dust record is available.
The paper is generally focused and straightforward, yet important, so it was a pleasure for me to review it.
I find the new dust record from U1537 very interesting. We can discuss if it is really better than ODP1090 (personnally I am quite convinced), but at the very least it is an alternative record which shows that ODP1090 should be taken with caution.
I do agree with Eric Wolff that the new borehole dust record from EPICA Dome C could get a bit more attention if it is really the first time such record is published. A quick comparison with the Coulter and lazer records would be interesting.
Regarding the end of the manuscript with the tuning strategy, I think there should be more powerful strategies, this is only a first-step strategy (but I think the authors are honnests in presenting it this way). And I do agree with Eric Wolff that this part is not as well presented and straightforward as the other parts. For example, the authors use a Nye ice flow model, but never explicitely describe the parameters they used, while for example the melting is a primary parameter for dating the old section of an ice core.
I would suggest to use the 1D model used in, e.g., Parrenin et al. (TC, 2017), Lilien et al. (TC, 2021) and Chung et al. (TC, in press), which uses a Lliboutry velocity profile. This model has an analytical thinning function and accounts for temporal variations of accumulation through a simple change of time variable. It should give a far better accuracy, while still being very easy to implement. I do not make a strong requirement to use this model, but I think it would be an improvement and I can provide guidance on request.
Moreover, the strategy is presented as decoupled between the modelling and the tuning, while I think both should be coupled: one first tune the top part, then apply the model with these dating constrains to extrapolate, then one tune the following part, etc. I personnally think the best approach would be to adjust the glaciological parameters of an ice core dating model in a Bayesian code like IceChrono/Paleochrono so as to optimize the fit with several targets, using a powerful MonteCarlo approach. (Such a method has been presented in the PhD of Jai Beeman in 2019, but unfortunately it has never been published elsewhere).
Minor comments:
- L. 40-45: The discussion of gradual vs abrupt MPT, as presented in Legrain et al. (2023) would fit nicely in this introduction, but I let you decide.
- L. ~50: In my opinion, the best evaluation of the age and state of basal ice in the Dome C area is from Chung et al. (TC, in press), but I let you decide if you want to cite it.
- L. 96: If I am correct, DFO2006 is the O2/N2 age scale of the first Dome Fuji ice core, DF1. The age scale from Dome Fuji members (2017) is DFO2006, then extended using AICC2012 for DF2. Please check, I don't think this age scale has a proper name, but I would call it DFO2006+AICC2012.
- L. ~120: In Table 1, the correlation coefficients are given for the ice core dust records, not for their log. The coeffs for the log-records are given a bit later in the manuscript, but not in Table 1. I personnally think the coeffs for the log-records are more relevant than for the raw records and I would put them in Table 1.
Citation: https://doi.org/10.5194/egusphere-2023-1342-RC2 -
AC3: 'Reply on RC2', Jessica Ng, 09 Nov 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1342/egusphere-2023-1342-AC3-supplement.pdf
-
RC3: 'Comment on egusphere-2023-1342', Lorraine Lisiecki, 02 Oct 2023
This manuscript is quite practically focused on initial assessment of drilling locations for oldest ice, proposing and demonstrating that borehole optical logs of dust would be an effective method to estimate the age of ice back to 1.5 Ma. Overall, I find this to be a valuable scientific contribution worthy of publication. However, a few portions of the manuscript need to be clarified and a few more calculations/experiments would also be very useful.
Needed clarifications:
- Line 117-118: The meaning here is ambiguous about what the authors did to “correlate” the records and put them “on a common timescale.” I think the intended meaning is that the records were sampled at common time steps in order to calculate the correlation coefficient between the signals. However, the phrasing could alternatively be interpreted as *stratigraphic* correlation that aligned the records to one common target timescale. Please clarify. If the records are all staying on their respective published timescales described in section 2, then a brief discussion of the extent to which these timescales are expected to agree or differ is warranted.
- Like Eric Wolff, I found Figure 5 to be extremely confusing. The caption, axes labels, and legend don’t seem to agree with one another. I think the blue line color was used for two different types of data. If so, the problem could be fixed by changing the line color for laser dust in panel c to another color that isn’t already used for something else.
- The description of how the artificial dust record was created is not clear, particularly “scaling the smoothed record by random factors between 0.4 - 0.7 linearly interpolated between 500 kyr intervals” (line 186). An additional concern I have about this methodology is that the laser dust record for EDC shows a much weaker dust amplitude from 500-800 ka than 0-500 ka. Are depth-dependent processes contributing to the weakening of the dust signal in the ice? If so, it would be more realistic if your artificial borehole dust record progressively decreased in amplitude farther back in time rather than scaling randomly through time.
- It is very difficult to visually discern misalignment in the DTW results, which strengthens the manuscript’s conclusion that DTW is not a suitable method for initial comparison between the borehole optical dust logs and the marine record. This point would be strengthened if the authors could find an additional way to illustrate the misalignment. For example, at the top of panel c, the artificial dust record could be plotted on its original, correct age model and arrows could be drawn to DTW results to show how dust peaks were misaligned.
Additional calculations/experiments:
- The weaker correlation for derivatives of the log of the data than the log itself is worthy of a bit more discussion/exploration. I wonder whether this might be an artifact of small differences in the age models for each record. The derivative has higher frequency variability and is therefore more likely to be affected by small age model differences/misalignments. This might actually make the derivative more useful for alignment than the log itself. The correlation of the derivative may also be advantageous if it is less sensitive to long-term trends in the mean or amplitude of the signal through time. This could potentially be evaluated by improving the alignment between the records or by analyzing synthetic records.
- Was the alignment experiment for an artificial dust signal repeated multiple times with different random noise or only once? Multiple iterations would be useful for characterizing the probability of incorrect conclusions. Additionally, what if the correlation between UF1537 and dust ice were slightly lower before 800 ka (e.g., R=0.7 instead of 0.76)?
- Eric Wolff makes a good point about the need to detect ice disturbances that could produce repeated or jumbled sections of the dust time series. I recommend creating an artificial signal that contains jumbled or duplicated sections (from the modified marine dust record) to determine whether your proposed method would detect a misfit between the simulated disturbed ice and the original marine dust record.
Citation: https://doi.org/10.5194/egusphere-2023-1342-RC3 -
AC2: 'Reply on RC3', Jessica Ng, 09 Nov 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1342/egusphere-2023-1342-AC2-supplement.pdf
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 316 | 112 | 28 | 456 | 25 | 25 |
- HTML: 316
- PDF: 112
- XML: 28
- Total: 456
- BibTeX: 25
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1